The Researcher's Need
A researcher needed a per-capita CO₂ emission variable for 34 Indonesian provinces over ten years (2015–2024) as input for a regression analysis. The source data is openly available — the global EDGAR emission database — but as a 0.1° grid per sector per year, not as provincial figures. Summing it correctly turns out to be far harder than it looks.
Three Pitfalls We Handled
- Boundary cells. A grid cell is often split between two provinces. We divide it by area share instead of assigning it whole to one province. For a small province such as Jakarta, other methods can shift the result by −18% to +60%.
- Coasts and sea. Emissions from coastal power plants, ports, and offshore activity are recorded in sea cells. We added a one-cell sea zone divided among the nearest provinces (Voronoi), while excluding sea that lies closer to neighbouring countries.
- Reconciliation. The 34 provinces must add up to the national total. Our final result is within 0.1% in every year.
How It Works
The whole process is written as one re-runnable Python pipeline: data download and validation, building the boundaries and coastal zone, area-share weights (cross-checked with two geometry engines), aggregation into 8 sector groups, division by population, and nine quality-control checks — from completeness and trend spikes to power-plant hotspots and double-counting.
Deliverables
- A 340-row province-year master dataset in Excel, CSV, and Stata, with variable metadata and a method comparison.
- A quality-control report and a cell-weight table for auditing.
- GeoPackage layers (land, coastal, and sea boundaries) ready for QGIS.
- Maps, charts, a presentation notebook, and a shareable HTML version.
An Honest Note
Emission data records emissions where they are produced. A single grid cell on the west coast of Banten holds almost half of Indonesia's 2024 power-sector emissions, which makes Banten look extremely high per capita. This is a property of the source data, not a processing error — and we reported it openly, along with the handling options, for the researcher to decide on before the variable is used.
Need Support for Your Research?
We support remote sensing, photogrammetry, spatial analysis, and grid-data research — see our Geospatial Research Support service. Pricing is negotiable; what matters is that your research gets finished, meets your expectations, and delivers its full value.
